{"id":"f667a4fe-1e42-4790-a76c-4db01a35ca52","arxiv_id":"2505.22080","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A network game shows that central trade countries concentrate their currency use, which amplifies the liquidity incentives of the issuer and solidifies dominant currencies.","lead":"This paper builds a game-theoretic model where trade networks and issuer liquidity provision reinforce each other in currency competition. It explains why the US dollar persists and why China's trade dominance has not made the renminbi a global currency.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The centrality mechanism rests on Lemma 1's linear Katz-Bonacich representation, which requires the quadratic cost; footnote 3 concedes the restriction, and no robustness check shows Proposition 2 survives nonlinear perturbations.","rationale":"In good faith, the model is internally coherent: existence (Lemma 2), the centrality representation (Lemma 1), and the issuer commitment incentives are derived consistently. The quadratic cost is the point where every subsequent result attaches. If the best-response map were nonlinear, the neat Katz-Bonacich statement would become an approximation and multiple equilibria could appear; Proposition 2's comparison of marginal responses and Proposition 4's Gini-centrality formula would need rederivation. The reader's weakest_assumption identifies the same spot, so I agree. I also noticed a smaller issue: Proposition 4's stated Gini formula is off by a factor of two (for T=2 the RHS is |d|/(2m) while equation (14) gives |d|/(4m)), and Lemma 3/Proposition 1 threshold proofs are sketched; these are secondary but support the conditional verdict. Therefore the verdict should remain CONDITIONAL, unchanged from the reader's recommendation, pending a robustness check of the quadratic specification and a tightening of the threshold proofs.","tokens_in":19222,"tokens_out":22166,"duration_ms":260668,"concrete_test":"Using the two-currency model of Section 2, fix a 5-node star network with w=0.25 on all links, m_i=1, β=2.2, f(e)=e^{0.5}, c(e)=exp(e), and add a small cubic perturbation ε·x_ip^3 to user i's cost. Solve for the SPE by backward induction—using the paper's entry and tie-breaking rules—for ε∈{0,0.001,0.01,0.1} and for the Proposition 2 preference asymmetry μ=1.5 assigned to the center versus a leaf. Record the sign and magnitude of e_a^*−e_b^* and X_a^*−X_b^*. If either sign reverses for small ε, the centrality mechanism is an artifact of quadratic costs; if it persists, the concern is mitigated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is Lemma 1 (Section 3.1), which writes the cross-currency usage difference as d = (I - w/β)^{-1}γ and identifies it with Katz-Bonacich centrality. This is exactly the linear best-response property of the quadratic cost C_ip(x_p,e_p) = (β/2)x_ip^2 - Σ_j w_ij x_ip x_jp - f_ip(e_p)x_ip; footnote 3 concedes the form is restrictive and justified only as a Taylor approximation. With a general convex cost, the FOC contains a nonlinear function of x_i, so d is no longer a linear resolvent of w. The feedback that 'central users prefer the currency of the issuer who commits more' (Proposition 2) relies on the additive decomposition of total usage difference D(e_a,e_b) in equation (17) and on linearity of the cross-partial comparison; both are artifacts of the quadratic specification. No robustness check or alternative functional form is offered, so the paper's central claim—network centrality, not trade volume, determines adoption—is supported only in the linear-quadratic case. This is a concern about external robustness rather than an internal contradiction; the proofs are coherent conditional on this functional form.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a three-stage game in which two currency issuers sequentially choose commitment levels for liquidity provision, and users in a weighted directed trade network allocate a fixed import volume across currencies to minimize a mean-variance cost. The central result is that cross-currency usage differences equal a Katz-Bonacich centrality weighted by liquidity preferences (Lemma 1), leading to predictions: the issuer of the currency preferred by more central users commits more and obtains a larger market share (Proposition 2); network integration strengthens the incumbent's deterrence (Proposition 3); and more central users concentrate their currency holdings, as measured by a Gini coefficient (Proposition 4). The two-currency model is extended to T currencies via a dynamic programming argument.","tokens_in":19497,"tokens_out":14175,"duration_ms":123464,"significance":"If the results hold, the paper offers a tractable theory of how network position, rather than aggregate trade volume, determines international currency adoption, and it generates testable predictions linking centrality to currency concentration. Strengths include closed-form solutions, a clean link to a standard network measure, and an explicit model of issuer commitment. The significance is reduced by the reliance on a linear-quadratic cost structure; the core mechanism is a consequence of linear best responses, and the paper does not demonstrate robustness to non-quadratic costs. Nonetheless, the model is self-contained and the proofs are largely coherent conditional on this structure.","major_comments":[{"comment":"The construction of the thresholds k and ¯k is not rigorous. The proof asserts monotonicity of ˆea(k) and that the two utility functions intersect at most twice, but no proof is given. The derivative expression for ∂[ua(ea,∅;k)−ua(˜ea,˜eb;k)]/∂k is not signed from the stated formula because ∂ea/∂k and ∂˜eb/∂k are unsigned. In the proof of Proposition 1, the threshold k appears before it is defined, and the comparative statics of k in β and M are argued using values that depend on the threshold itself. These gaps are load-bearing for Proposition 1, which is the paper's main characterization of the incumbent advantage.","section":"Appendix, Proof of Lemma 3 and Proof of Proposition 1"},{"comment":"The central mechanism is derived from the quadratic cost function, which yields the linear best-response mapping d = (I − w/β)^{-1}γ. Footnote 3 concedes the restriction, but no robustness check shows that the results of Propositions 2 and 4 survive non-quadratic perturbations. For example, adding a small cubic term to the cost would break the linear resolvent representation, and it is not shown that the ranking in Proposition 2 is locally robust. Since the paper claims that network centrality, not trade volume, determines adoption as a general insight, the authors should either provide a perturbation argument or explicitly restrict the claims to the quadratic case.","section":"Section 3.1 and Section 4.2, Eq. (7) and Eq. (17)"},{"comment":"The proof uses the identity x_iτ − x_i,τ+c = Σ d_it, but the subsequent reduction to a sum of t(T−t)|d_it| requires that all d_it have the same sign, i.e., that the currencies are ordered by decreasing usage. Without this ordering, the formula is false. For T=3 and allocations (0.2, 0.5, 0.3), Eq. (14) gives G_i=0.2, while the claimed sum gives 0.333. Because the Gini coefficient is invariant to reordering, the result can be repaired, but as written the statement and its proof are incorrect.","section":"Section 5.2, Proof of Proposition 4"},{"comment":"The condition guaranteeing interiority and order-independence of the T-currency solution is not proved. The condition mi,t ≥ sup{m ∈ (0, mi0]: C'_it(m) ≤ C'_{i,t+1}(0)} involves the endogenous state variable mi,t, and it is not shown that it holds for the β ≥ β region. The claim that low β makes the order of currencies matter conflicts with the convexity of the allocation problem, which should yield a unique optimum regardless of order. Since Corollary 1 and Proposition 4 rely on this solution, the T-currency extension is not yet fully rigorous.","section":"Section 5.1, dynamic programming extension"}],"minor_comments":[{"comment":"The term 'Kat-Bonacich' should be 'Katz-Bonacich'.","section":"Lemma 1"},{"comment":"The notation f_t as a vector difference is not defined, and the quantity γt is introduced as 1/β [ft − ft+1] but the subscript t is used both for currency and stage.","section":"Section 5.1, Eq. (12)"},{"comment":"The factor in the derivative after Eq. (17) appears to be off by a factor of 2; the correct expression should be (µ−1)(κi−κj) f'(e)/(2β). The conclusion is unaffected since the sign is correct.","section":"Proposition 2 proof, after Eq. (17)"},{"comment":"The reference to Kikuchi et al. (2025) lists the university as 'Cornel University'; it should be 'Cornell University'.","section":"References"},{"comment":"The captions of Figures 2 and 4 do not explain what is plotted on the axes, beyond referring to commitment levels and k.","section":"Figures 2 and 4"},{"comment":"The symbol C_{ip} is used first for the unit cost and then for the total perceived cost; this should be clarified to avoid confusion.","section":"Section 2, cost notation"}],"recommendation":"major_revision","confidential_remarks":"The paper's core idea is interesting and the model is self-contained, but the proof of Proposition 1 and Lemma 3 lacks rigor, and Proposition 4 is incorrect as stated without an ordering assumption. The quadratic cost restriction is acknowledged but not treated as a limitation in the paper's claims. These issues are fixable with additional analysis, so I recommend major revision rather than rejection. The editor may also wish to verify the novelty relative to Belhaj and Deroïan (2014) and Chen et al. (2018, 2022), which cover multi-action network games with resource allocation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a genuinely useful theory paper. It formalizes the intuition that a country's position in the trade network, not its aggregate trade volume, drives currency adoption, and it shows how that feeds back into issuers' liquidity decisions. The Katz-Bonacich representation (Lemma 1) is the right tool, and the extension to T currencies via dynamic programming is a real addition. I'd send it to a referee.\n\nWhat's new: previous work on dominant currencies often either uses aggregate network externalities or ignores the strategic supply side. This paper puts both in one game and gets clean comparative statics: Propositions 2 and 3 connect centrality to market share and show network integration entrenches the incumbent. The T-currency section is more than an afterthought—it derives a Gini coefficient that is literally a sum of adjusted centralities (Prop 4), which gives a testable cross-country prediction.\n\nSoft spots: the proofs, especially Lemma 3 and Proposition 1, are sketched. The existence and monotonicity arguments rely on unstated crossing conditions; a referee will need to see those filled in. The bigger issue is the quadratic cost assumption. The stress-test is right that Lemma 1's linear resolvent is exactly the linear best-response property, and footnote 3 concedes the form is restrictive. Without a robustness check under a nonlinear cost, the strong claim that centrality, not volume, determines adoption is only shown for this functional form. That's a real limitation, but it's the standard kind of tractability assumption in network games, and the paper frames it honestly. I don't think it's a fatal flaw; it limits external validity, not internal consistency. Also, the empirical illustrations are just centralities from UN Comtrade—no formal test. That's fine for a theory paper, but it should be advertised as motivation, not evidence.\n\nWho it's for: international macro and trade people interested in currency internationalization, and network game theorists. It deserves a serious referee. My recommendation: engage with it, ask for tightened proofs and at least one robustness exercise on the cost function, and it should be publishable.","headline":"Solid theory paper on how trade network centrality, not trade volume, drives currency adoption; core mechanism is transparent, but proofs are sketched and the linear-quadratic cost is a real limitation.","tokens_in":19960,"tokens_out":3566,"would_cite":false,"duration_ms":37372,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that a currency's fate is set by the network centrality of the users who prefer it, not by the issuer's trade volume, through an adoption–liquidity feedback loop.","keywords":["dominant currency","trade network","Katz-Bonacich centrality","liquidity provision","currency internationalization","network spillover","subgame perfect equilibrium","US-China competition"],"falsifier":"Compute each country's settlement-currency Gini coefficient and its Katz-Bonacich centrality in bilateral trade-payment data; if the relationship is flat, or if increasing network integration does not reduce the number of currencies used, the model's core prediction is falsified.","tokens_in":19032,"feed_emoji":"💱","tokens_out":7278,"duration_ms":72492,"temperature":0.7,"pith_summary":"The paper tries to show that a currency becomes dominant not because its issuer trades the most, but because the countries sitting at the center of the trade network prefer it. It builds a three-stage game in which currency issuers first commit to providing liquidity, then users choose which currencies to settle trade in, and those choices feed back into issuers' incentives. The central finding is that a user's currency allocation is proportional to its network centrality, and the issuer preferred by the most central users commits more and wins the largest market share. This explains why a large aggregate trade share, as China has, does not automatically produce a global currency, and why a more integrated trade network tends to entrench the incumbent currency.","feed_headline":"Centrality, not trade volume, decides which currency wins","feed_subtitle":"Network position, not trade volume, drives currency dominance through an adoption–liquidity feedback loop.","key_machinery":"The central object is the Katz-Bonacich centrality vector $\\kappa = (I - \\lambda w)^{-1}\\mathbf{1}$, which counts all directed paths in the trade-payment network with decay factor $\\lambda$; the paper's user equilibrium sets $\\lambda = 1/\\beta$ and replaces the vector of ones with the marginal-cost-difference vector $\\gamma$, making usage differences an adjusted centrality. This linear resolvent identity converts network spillovers into a closed-form mapping from issuer commitments to currency shares, and the same identity, read through a Bellman equation, carries the two-currency result to a setting with finitely many currencies.","core_discovery":"On the paper's own terms, the equilibrium usage of currencies in a trade network is governed by a feedback loop between users and issuers, and the decisive variable is network position rather than aggregate trade volume. Formally, the usage difference $d_i = x_{ia} - x_{ib}$ satisfies $d = (I - \\beta^{-1}w)^{-1}\\gamma$, so each user's currency bias is an adjusted Katz-Bonacich centrality weighted by the marginal-cost advantage $\\gamma_i = (f_{ia}(e_a) - f_{ib}(e_b))/\\beta$. A more central user's choice reaches more trading partners, so that user concentrates on a single currency more strongly; the issuer whose currency it prefers therefore commits more liquidity and captures a larger market share. When users are symmetric, network integration amplifies the incumbent's commitment advantage and reduces the number of international currencies in equilibrium. The Gini coefficient of a user's currency allocation is a sum of adjusted Katz-Bonacich centralities, so central countries concentrate their settlement usage while peripheral countries diversify.","pith_inferences":["Beyond the paper, the Gini-centrality link suggests an immediate empirical design: compute each country's settlement-currency Gini coefficient from payments data and regress it on Katz-Bonacich centrality in bilateral trade-payment flows, expecting a positive cross-sectional slope.","Beyond the paper, the model implies that capital-account liberalization is endogenous to network position, so cross-country studies of currency internationalization should treat liberalization as a strategic choice rather than an exogenous policy variable.","Beyond the paper, the sequential timing is what selects the unique incumbent-deterrence outcome; under simultaneous entry the same model has multiple equilibria, so the theory does not by itself rule out a multipolar outcome under simultaneous competition."],"forward_implications":["A country's settlement-currency concentration should track its Katz-Bonacich centrality in the trade-payment network, with the most central countries showing the most skewed currency portfolios.","Two currencies can coexist when the cost of providing liquidity is low relative to total demand $M$, and even more so when exchange-rate risk $\\beta$ is high, because users want to diversify.","Deepening trade integration lowers the threshold at which the incumbent deters entry, so a more connected world should see fewer settlement currencies.","An emerging currency can overtake the incumbent only when network restructuring raises the centrality of a user that prefers it; otherwise the incumbent's first-mover liquidity commitment blocks entry.","If capital-account openness is itself a commitment to liquidity, then a country's network position endogenously determines how open it will be, making aggregate trade dominance a poor predictor of currency internationalization."],"supporting_citations":[{"why":"Supplies the network-game framework with strategic complementarities whose linear best responses produce the centrality-based usage solution.","marker":"Ballester et al. (2006)"},{"why":"Defines the status-index centrality whose infinite-path series appears as the resolvent in Lemma 1.","marker":"Katz (1953)"},{"why":"Introduces the family of centrality measures used to interpret the usage-difference vector.","marker":"Bonacich (1987)"},{"why":"Characterizes equilibrium when agents allocate fixed resources between two activities, the closest prior network-allocation result the paper generalizes to $T$ currencies.","marker":"Belhaj and Deroïan (2014)"},{"why":"Motivates the modeling choice that currency internationalization begins with trade invoicing and settlement.","marker":"Eichengreen (2011)"}],"fun_headline_variants":["In trade webs, centrality beats trade volume for currency dominance","The centrality feedback loop that crowns a single dominant currency","Network position, not trade size, decides the winning currency","How central users pick one currency and make it dominant"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes users' cost functions are exactly quadratic (mean-variance preferences), which makes their best responses linear and turns currency concentration into a centrality transform; if actual settlement costs are not quadratic, the proportionality results can fail.","fun_headline_variants_meta":{"raw":{"variants":["In trade webs, centrality beats trade volume for currency dominance","The centrality feedback loop that crowns a single dominant currency","Network position, not trade size, decides the winning currency","How central users pick one currency and make it dominant"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000194,"raw_usage":{"total_tokens":1302,"prompt_tokens":839,"completion_tokens":463,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":398}},"tokens_in":455,"tokens_out":463,"duration_ms":5370,"temperature":1.0,"reasoning_tokens":398,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T13:15:46.716385+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute each country's settlement-currency Gini coefficient and its Katz-Bonacich centrality in bilateral trade-payment data; if the relationship is flat, or if increasing network integration does not reduce the number of currencies used, the model's core prediction is falsified.","supporting_citations":[{"cited_title":"(1953): A new status index derived from sociometric analysis, Psychometrika, 18, 39--43","cited_arxiv_id":null,"evidence_quote":"Defines the status-index centrality whose infinite-path series appears as the resolvent in Lemma 1."},{"cited_title":"(1987): Power and centrality: A family of measures, American Journal of Sociology, 92, 1170--1182","cited_arxiv_id":null,"evidence_quote":"Introduces the family of centrality measures used to interpret the usage-difference vector."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Characterizes equilibrium when agents allocate fixed resources between two activities, the closest prior network-allocation result the paper generalizes to $T$ currencies."},{"cited_title":"(2011): The renminbi as an international currency, Journal of Policy Modeling, 33, 723--730","cited_arxiv_id":null,"evidence_quote":"Motivates the modeling choice that currency internationalization begins with trade invoicing and settlement."}],"review_version":1}